Latest AI Headlines

OpenAI has launched deep research in ChatGPT, a feature that conducts multi-step online research tasks. The tool searches, analyzes, and synthesizes information from various web sources to create detailed reports. Currently available to Pro users, with planned rollout to Plus and Team users.

OpenAI has released o3-mini 🧠, a cost-efficient STEM reasoning model available in ChatGPT and the API. It improves on o1-mini with better accuracy, speed, and lower latency. Developers can adjust reasoning effort (low, medium, high) and use function calling, structured outputs, and developer messages.

OpenAI released Operator as a research preview, an AI agent 🤖 that can perform web tasks by typing, clicking, and scrolling through its own browser interface. Currently available to Pro users in the U.S., this preview allows OpenAI to learn from user feedback while refining the technology.

DeepSeek, a Chinese AI startup 🇨🇳, has released DeepSeek-R1, an open-source reasoning language model that matches the performance of OpenAI's o1 model across math, coding, and reasoning tasks. The model is available at 90-95% lower cost than o1 and is accessible through both open-source releases and an API.

ChatGPT has a new feature called tasks ✅ that lets you plan things ahead of time 🗓️. You can now tell ChatGPT to do something for you later, either just once or over and over again on a schedule.

Cohere has entered the intense AI agents arms race 🏃‍♂️ with North, a new workplace platform that helps businesses use AI in their everyday operations.

NVIDIA has released Cosmos 🪐, a platform featuring world foundation models, tokenizers, and data processing tools designed for developing autonomous vehicles and 🦾 robots.

NVIDIA has launched new AI Blueprints focused on building AI agents 🤖 - software that can reason, plan, and take action. Through partnerships with CrewAI, Daily, LangChain, LlamaIndex, and Weights & Biases, these agents can analyze large amounts of data, create summaries from various content types, and process information in real-time.

OpenAI announced o3 and o3-mini, new AI reasoning models that use "private chain of thought" to plan responses. While not yet publicly released, they will be available for safety testing and research. The o3 model achieved notable scores in mathematics and scientific reasoning benchmarks. 🚀

ElevenLabs launched Flash, a new speech synthesis 🗣️ model with 75ms latency, moving into a territory where AI voice interactions start feeling more human. 🙊

Google has launched an experimental AI model called Gemini 2.0 Flash Thinking that shows its step-by-step reasoning process 🧠 while solving problems. This builds on the recently released Gemini Flash 2.0 platform and competes with OpenAI's o1 reasoning model.

NVIDIA launches a compact generative AI supercomputer, the Jetson Orin Nano Super Developer Kit, at $249 (reduced from $499). The palm-sized device offers up to 1.7x better generative AI performance compared to its predecessor, targeting developers, hobbyists, and students.

OpenAI has released new API features including WebRTC support for voice applications 🗣️, significant cost reductions for audio processing, and improvements to model fine-tuning. These updates simplify voice AI development and make it more cost-effective.

Google DeepMind has announced Veo 2, a new AI video generation model 📹 that can create two-minute videos in 4K resolution. This is currently only available through Google's VideoFX tool.

Microsoft has released Phi-4, a 14B parameter small language model (SLM) that matches or exceeds larger models' performance in complex reasoning and math tasks.

Google introduces Agentspace, a new enterprise platform that combines Gemini AI with Google search capabilities and enterprise data integration. This solution aims to solve the problem of scattered organizational knowledge 🏢, where employees typically need 4-6 different tools to find information and complete tasks.

Google is updating NotebookLM with a redesigned interface, interactive audio features, and launching NotebookLM Plus, a premium subscription service.

Google DeepMind has released Gemini 2.0, including an experimental version called Gemini 2.0 Flash. The release places Google even closer to OpenAI and Anthropic in the race to have the most capable AI chat.

OpenAI is releasing Sora Turbo, an upgraded version of their text-to-video 📹 AI model, to ChatGPT Plus and Pro users through Sora.com. This release moves the technology from research preview to a standalone product, offering higher resolution capabilities and longer video durations.

Anthropic has released a new version of Claude 3.5 Sonnet that can directly interact with computer interfaces by moving cursors, clicking, and typing - similar to how humans use computers. This capability is currently in public beta and represents a shift from AI models using specialized tools to being able to work with standard computer software.